🥤 AI Eats 💩 and Preview of AI Regulation Framework

It’s the Soda Road. The only newsletter that you can recommend to a friend without getting blocked. Help a friend get smarter about AI. Share the Soda.

Here’s what your friends wish they knew today:

• AI is dealing drugs again. Not quite like Walter White, but 💩 is involved.

• One small step for man, one giant leap for AI regulation. A sneak preview of how Governments may approach AI.

DRUG DEALING AI

A few weeks ago we told you about DALL-E creating models of disease-fighting proteins. We left you on a cliffhanger, waiting to hear what happens next... Well, the next chapter of the manga is finally here!

Some nerds scientists created an AI (similar to ChatGPT) and trained it on the structure of 280 million amino acids. They created a model that takes a group of proteins and generates a new similar structure.

Of the initial 100 test molecules, 66 of them were similar to proteins that destroy bacteria. 2 of the molecules were even able to kill E. coli bacteria (the bacteria found in poo poo 💩).

TL;DR - You can use AI to create a shit-eating protein.

ROBO-POCALYPSE AVERTED: GUIDELINES FOR AI

The National Institute of Standards and Technology (NSIT) released its Artificial Intelligence Risk Management Framework (AI RMF 1.0). This was developed with Government, private, and public businesses.

Governments could use the AI RMF as a guideline for enforcing AI practices in companies. Uncle Sam works in sneaky ways 👀

The NSIT published a simple playbook that breaks down each of the 4 main components of their AI Risk Management Framework. We’ve explained them in lameman’s terms here.

Map

The MAP function identifies the potential risks of using an AI system. Framing these risks within the specific use cases of the AI.

Without a proper understanding of the context and risks, it’s hard to manage and prevent potential problems. The MAP function allows a company to identify risks and their underlying factors.

The information gathered through the MAP function is used in the MEASURE and MANAGE functions. Helping to evaluate and manage the risks.

Measure

The MEASURE function evaluates, analyzes, compares, and tracks the risks and impacts associated with AI.

It uses the information gathered from the MAP function to guide the risk monitoring and management of the MANAGE function.

Manage

The MANAGE function handles the risks identified through the previous functions. It reduces the risks of system failure and negative impacts by using the data from the other functions.

The process of risk treatment includes making plans to respond to, recover from, and communicate about incidents or events.

There should be plans to prioritize risks and continuously monitor and improve the system's safety.

Govern

The GOVERN function is an ongoing and important aspect of managing the risks associated with using an AI system. It's required through the entire lifespan of the AI system and applies to the entire organization. It helps to enable the other four functions of AI risk management.

The outcomes of this function create a culture of risk management within the organization. Involved in designing, developing, or deploying AI systems.

The categories in this function interact with each other and with the other functions but don't depend on previous actions.

TL;DR

Companies need to understand what they’re getting themselves into when using AI, how to reduce risks, and how to recover from incidents. Or else Uncle Sam will make you eat E. coli.

That's everything.

Send us cool AI stuff on Twitter – we're always looking for new stories. Until then ✌🏼